ABSTRACT The global AIDS epidemic is one of the greatest threats to human health and development. Countries need to ground their AIDS strategies in an understanding of their own epidemics and their national re- sponses. A reliable estimation and prediction on the HIV/AIDS epidemic can help policy makers and program planners ef?ciently allocate the resource, plan and manage the intervention, treatment and care programs, evaluate their effort, and raise funds. The epidemiological model called ?EPP? has been used by the UNAIDS and most countries in the world for estimation and short-term prediction of HIV/AIDS trends from limited surveillance data since 2001. As the epidemic enters its fourth decade, both the epidemic and monitoring data sources have changed signi?cantly. The existing models are no longer capable of capturing the epidemic trend accu- rately. In this proposal, we plan to develop new statistical models to estimate and project the epidemic that re?ect those changes and utilize new data sources, with the following aims. 1. Incorporate new sources of data in current mathematical models to strengthen the epidemic esti- mates for high HIV prevalence countries (known as generalized epidemic). 2. Develop hierarchical models to provide reliable estimates of HIV epidemics for sub-national areas and key populations who are at higher risk for HIV, based on sexual practices, occupations, and substance use. 3. Develop Bayesian models to improve the estimation of key population sizes. 4. We will produce freely-available, open-source software to implement all of the proposed models and methods.